Design And Implementation Of Personalized Music Recommendation System Based On Sentiment Analysi

With the advancement of technology,the number of online music consumers is increasing,and people’s demands for music are also increasing.Therefore,music recommendation systems,as intelligent music recommendation technologies,have attracted the attention of music enthusiasts and industry professionals.The study of music recommendation systems is of great significance.Firstly,it can improve the accuracy and efficiency of music recommendation,enabling users to quickly find music they like.Secondly,music recommendation systems can provide powerful assistance for musicians and music companies,promoting the development of the music industry.However,traditional recommendation methods for music software suffer from the problems of low recommendation accuracy and lack of personalization,and the large-scale data also leads to difficulties in data storage,low computational efficiency,cold start,and data sparsity.These problems affect the accuracy of recommendation results,thus affecting users’ experience.To improve user experience and enhance recommendation efficiency,this thesis conducts research on the problems of traditional recommendation methods and proposes a mixed recommendation method that introduces emotional weights based on sentiment analysis of user comments.The thesis achieves personalized music recommendation based on sentiment analysis.Firstly,the development and relevant theories of music recommendation systems are summarized.Then,the required technologies are studied in depth,and the system modules are analyzed and designed accordingly,leading to the implementation of a complete music recommendation system.The main contributions of this thesis include:(1)The implementation of a sentiment analysis process based on song comments.The web page source code of the online music platform is analyzed,and information is collected using a third-party library based on Python.The Snow NLP library based on Naive Bayes is used for sentiment analysis,and the included Bayesian model is trained using a corpus and sentiment dictionary specifically for song comments.The thesis improves the sentiment analysis process for song comments and demonstrates through experiments that the sentiment analysis model has more accurate results.(2)The implementation of a mixed recommendation algorithm that introduces emotional weights.To solve the cold start problem,the sentiment analysis results of song comments are used as the song’s own characteristics for content-based recommendation,which is performed before constructing the rating matrix.The thesis also implements a weighted mixed recommendation based on both sentiment attributes and collaborative filtering,achieving a sentiment-based mixed recommendation.Finally,the thesis designs experiments and evaluates the algorithm using three evaluation metrics for music recommendation systems,proving that the algorithm has better recommendation performance.(3)The design and implementation of a music recommendation system.The sentiment analysis and recommendation processes are designed and implemented in Python using the Py Charm development platform.The system adopts a front-end and back-end separation structure,with the back-end developed using the Spring Boot framework based on Java and the front-end developed using the Vue framework.Finally,the system is tested to verify its feasibility and stability.